Connecting Copilot Studio with Power Automate | Day 102
Connect Copilot Studio to Power Automate agent flows to turn conversations into secure, reliable business actions.
- Published
- Reading time
- 3 min read
Quick answer
A Copilot Studio agent can pass inputs to a Power Automate agent flow, invoke business actions and use the returned outputs to respond to the user.

Week 15 · Day 102 of 365 in 365 Days of Copilot Studio — view the full series
Before you start
Is this guide for you?
- Best entry point
- 365 Days of Copilot Studio
- Time investment
- 3 min read
- Prerequisite knowledge
- An existing Copilot Studio agent and familiarity with Power Automate and the target business system.
What you’ll learn
- Agent flows connect conversations to business actions.
- Clearly named inputs and outputs help the agent use a flow.
- A submission confirmation does not mean an approval has been granted.
- Validate inputs, secure connections and handle failures.
Before you start
- Prerequisite knowledge
- An existing Copilot Studio agent and familiarity with Power Automate and the target business system.
On this page (6 sections)
Day 102 of 365 Days of Copilot Studio builds on Day 101 by connecting an agent to Power Automate. The goal is to turn a conversation into a business action and return a useful result.
Why connect an agent to a flow?
Copilot Studio handles the conversation; Power Automate can coordinate the steps behind it. Depending on the configured connectors and permissions, a flow can retrieve information, create or update records, start an approval or send a notification. This can reduce manual follow-up while keeping the process connected to its business systems.
Build your first agent flow
- Open your agent in Copilot Studio and navigate to Tools.
- Add a new tool and create an agent flow. Define the inputs the agent will provide, using clear names and descriptions.
- Add the business actions and configure their connections. Validate required values before acting.
- Use the flow's response action to return the outputs the agent needs.
- Save and publish the flow, add or configure it for the agent as needed, and test the complete conversation.
The exact interface can change. Check that the flow uses the supported When an agent calls the flow trigger and Respond to the agent action, and that the flow and agent are in the appropriate environment.
Pass inputs and return outputs
Consider an employee information lookup. The agent supplies an employee ID; the flow queries an authorized data source, finds the matching record and returns selected fields such as name, department and status. The agent can then present the result in natural language.
Define what happens when the ID is missing, the record is not found or the service is unavailable. Return only the data needed for the task and verify that the user and connection are authorized to access it.
Example: an employee leave request
- The agent collects the employee's leave dates and other required details and confirms the request.
- The flow validates the inputs, records the request in an approved system such as SharePoint or Dataverse, and starts the approval process.
- A notification can alert the responsible manager through Teams or Outlook.
- The flow returns a submission result and reference to the agent.
Submission is not approval. The agent should report that the request was submitted, and communicate an approval decision only when the workflow confirms it. For an approval that takes longer than the agent's response window, use a separate notification or status lookup rather than waiting in the conversation.
How the pieces work together
A user asks for a ticket status, the agent selects an appropriate tool and passes the ticket identifier to the flow. The flow retrieves authorized data, returns a structured result, and the agent communicates it to the user. Power Automate connectors can also support actions across SharePoint, Teams, Outlook, Dataverse and other services; availability and licensing depend on the connector and environment.
Five practices for reliable automation
- Keep flows focused: give each flow a clear business purpose and descriptive inputs and outputs.
- Validate inputs: check required values and formats before executing actions.
- Secure connections: use appropriate authentication, least-privilege access and data policies.
- Handle errors: return meaningful success and failure results without exposing sensitive details.
- Plan execution time: account for connector limits and the supported response timeout; use a separate follow-up for long-running work.
Test normal requests, missing or invalid inputs, unauthorized access and service failures with representative users. Review the flow's run history and assign an owner for monitoring and maintenance. Day 103 continues with passing parameters between Copilot Studio and Power Automate.
Tagged
AI Agents · Automation · Governance
Frequently asked questions
What can an agent flow do?
A flow can perform supported actions such as retrieving records, creating requests or sending notifications, subject to its connections and permissions.
How does the agent receive a result?
Configure the flow to return structured outputs using its response action, then use those outputs in the agent conversation.
Does submitting a request mean it is approved?
No. Confirm submission separately and communicate approval only after the approval workflow has actually confirmed it.
Have a Microsoft 365 topic idea?
Share article suggestions, community session ideas, corrections, or real-world scenarios for future nextM365 learning notes.
Keep learning Microsoft 365
Explore more practical guides for SharePoint, Power Platform, Copilot Studio, migration, automation, governance, and security.
Continue learning
Related tutorials
Related questions
Related comparisons
Next action